You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

432 lines
15KB

  1. /*
  2. * AVS encoding using the xavs library
  3. * Copyright (C) 2010 Amanda, Y.N. Wu <amanda11192003@gmail.com>
  4. *
  5. * This file is part of FFmpeg.
  6. *
  7. * FFmpeg is free software; you can redistribute it and/or
  8. * modify it under the terms of the GNU Lesser General Public
  9. * License as published by the Free Software Foundation; either
  10. * version 2.1 of the License, or (at your option) any later version.
  11. *
  12. * FFmpeg is distributed in the hope that it will be useful,
  13. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  15. * Lesser General Public License for more details.
  16. *
  17. * You should have received a copy of the GNU Lesser General Public
  18. * License along with FFmpeg; if not, write to the Free Software
  19. * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20. */
  21. #include <stdio.h>
  22. #include <stdlib.h>
  23. #include <string.h>
  24. #include <math.h>
  25. #include <stdint.h>
  26. #include <float.h>
  27. #include <xavs.h>
  28. #include "avcodec.h"
  29. #include "internal.h"
  30. #include "libavutil/opt.h"
  31. #define END_OF_STREAM 0x001
  32. #define XAVS_PART_I8X8 0x002 /* Analyze i8x8 (requires 8x8 transform) */
  33. #define XAVS_PART_P8X8 0x010 /* Analyze p16x8, p8x16 and p8x8 */
  34. #define XAVS_PART_B8X8 0x100 /* Analyze b16x8, b*/
  35. typedef struct XavsContext {
  36. AVClass *class;
  37. xavs_param_t params;
  38. xavs_t *enc;
  39. xavs_picture_t pic;
  40. uint8_t *sei;
  41. int sei_size;
  42. AVFrame out_pic;
  43. int end_of_stream;
  44. float crf;
  45. int cqp;
  46. int b_bias;
  47. float cplxblur;
  48. int direct_pred;
  49. int aud;
  50. int fast_pskip;
  51. int mbtree;
  52. int mixed_refs;
  53. int64_t *pts_buffer;
  54. int out_frame_count;
  55. } XavsContext;
  56. static void XAVS_log(void *p, int level, const char *fmt, va_list args)
  57. {
  58. static const int level_map[] = {
  59. [XAVS_LOG_ERROR] = AV_LOG_ERROR,
  60. [XAVS_LOG_WARNING] = AV_LOG_WARNING,
  61. [XAVS_LOG_INFO] = AV_LOG_INFO,
  62. [XAVS_LOG_DEBUG] = AV_LOG_DEBUG
  63. };
  64. if (level < 0 || level > XAVS_LOG_DEBUG)
  65. return;
  66. av_vlog(p, level_map[level], fmt, args);
  67. }
  68. static int encode_nals(AVCodecContext *ctx, AVPacket *pkt,
  69. xavs_nal_t *nals, int nnal)
  70. {
  71. XavsContext *x4 = ctx->priv_data;
  72. uint8_t *p;
  73. int i, s, ret, size = x4->sei_size + FF_MIN_BUFFER_SIZE;
  74. if (!nnal)
  75. return 0;
  76. for (i = 0; i < nnal; i++)
  77. size += nals[i].i_payload;
  78. if ((ret = ff_alloc_packet2(ctx, pkt, size)) < 0) {
  79. return ret;
  80. }
  81. p = pkt->data;
  82. /* Write the SEI as part of the first frame. */
  83. if (x4->sei_size > 0 && nnal > 0) {
  84. memcpy(p, x4->sei, x4->sei_size);
  85. p += x4->sei_size;
  86. x4->sei_size = 0;
  87. }
  88. for (i = 0; i < nnal; i++) {
  89. s = xavs_nal_encode(p, &size, 1, nals + i);
  90. if (s < 0)
  91. return -1;
  92. p += s;
  93. }
  94. pkt->size = p - pkt->data;
  95. return 1;
  96. }
  97. static int XAVS_frame(AVCodecContext *ctx, AVPacket *pkt,
  98. const AVFrame *frame, int *got_packet)
  99. {
  100. XavsContext *x4 = ctx->priv_data;
  101. xavs_nal_t *nal;
  102. int nnal, i, ret;
  103. xavs_picture_t pic_out;
  104. x4->pic.img.i_csp = XAVS_CSP_I420;
  105. x4->pic.img.i_plane = 3;
  106. if (frame) {
  107. for (i = 0; i < 3; i++) {
  108. x4->pic.img.plane[i] = frame->data[i];
  109. x4->pic.img.i_stride[i] = frame->linesize[i];
  110. }
  111. x4->pic.i_pts = frame->pts;
  112. x4->pic.i_type = XAVS_TYPE_AUTO;
  113. x4->pts_buffer[ctx->frame_number % (ctx->max_b_frames+1)] = frame->pts;
  114. }
  115. if (xavs_encoder_encode(x4->enc, &nal, &nnal,
  116. frame? &x4->pic: NULL, &pic_out) < 0)
  117. return -1;
  118. ret = encode_nals(ctx, pkt, nal, nnal);
  119. if (ret < 0)
  120. return -1;
  121. if (!ret) {
  122. if (!frame && !(x4->end_of_stream)) {
  123. if ((ret = ff_alloc_packet2(ctx, pkt, 4)) < 0)
  124. return ret;
  125. pkt->data[0] = 0x0;
  126. pkt->data[1] = 0x0;
  127. pkt->data[2] = 0x01;
  128. pkt->data[3] = 0xb1;
  129. pkt->dts = 2*x4->pts_buffer[(x4->out_frame_count-1)%(ctx->max_b_frames+1)] -
  130. x4->pts_buffer[(x4->out_frame_count-2)%(ctx->max_b_frames+1)];
  131. x4->end_of_stream = END_OF_STREAM;
  132. *got_packet = 1;
  133. }
  134. return 0;
  135. }
  136. x4->out_pic.pts = pic_out.i_pts;
  137. pkt->pts = pic_out.i_pts;
  138. if (ctx->has_b_frames) {
  139. if (!x4->out_frame_count)
  140. pkt->dts = pkt->pts - (x4->pts_buffer[1] - x4->pts_buffer[0]);
  141. else
  142. pkt->dts = x4->pts_buffer[(x4->out_frame_count-1)%(ctx->max_b_frames+1)];
  143. } else
  144. pkt->dts = pkt->pts;
  145. switch (pic_out.i_type) {
  146. case XAVS_TYPE_IDR:
  147. case XAVS_TYPE_I:
  148. x4->out_pic.pict_type = AV_PICTURE_TYPE_I;
  149. break;
  150. case XAVS_TYPE_P:
  151. x4->out_pic.pict_type = AV_PICTURE_TYPE_P;
  152. break;
  153. case XAVS_TYPE_B:
  154. case XAVS_TYPE_BREF:
  155. x4->out_pic.pict_type = AV_PICTURE_TYPE_B;
  156. break;
  157. }
  158. /* There is no IDR frame in AVS JiZhun */
  159. /* Sequence header is used as a flag */
  160. if (pic_out.i_type == XAVS_TYPE_I) {
  161. x4->out_pic.key_frame = 1;
  162. pkt->flags |= AV_PKT_FLAG_KEY;
  163. }
  164. x4->out_pic.quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
  165. x4->out_frame_count++;
  166. *got_packet = ret;
  167. return 0;
  168. }
  169. static av_cold int XAVS_close(AVCodecContext *avctx)
  170. {
  171. XavsContext *x4 = avctx->priv_data;
  172. av_freep(&avctx->extradata);
  173. av_free(x4->sei);
  174. av_freep(&x4->pts_buffer);
  175. if (x4->enc)
  176. xavs_encoder_close(x4->enc);
  177. return 0;
  178. }
  179. static av_cold int XAVS_init(AVCodecContext *avctx)
  180. {
  181. XavsContext *x4 = avctx->priv_data;
  182. x4->sei_size = 0;
  183. xavs_param_default(&x4->params);
  184. x4->params.pf_log = XAVS_log;
  185. x4->params.p_log_private = avctx;
  186. x4->params.i_keyint_max = avctx->gop_size;
  187. if (avctx->bit_rate) {
  188. x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
  189. x4->params.rc.i_rc_method = XAVS_RC_ABR;
  190. }
  191. x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
  192. x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
  193. x4->params.rc.b_stat_write = avctx->flags & CODEC_FLAG_PASS1;
  194. if (avctx->flags & CODEC_FLAG_PASS2) {
  195. x4->params.rc.b_stat_read = 1;
  196. } else {
  197. if (x4->crf >= 0) {
  198. x4->params.rc.i_rc_method = XAVS_RC_CRF;
  199. x4->params.rc.f_rf_constant = x4->crf;
  200. } else if (x4->cqp >= 0) {
  201. x4->params.rc.i_rc_method = XAVS_RC_CQP;
  202. x4->params.rc.i_qp_constant = x4->cqp;
  203. }
  204. }
  205. if (x4->aud >= 0)
  206. x4->params.b_aud = x4->aud;
  207. if (x4->mbtree >= 0)
  208. x4->params.rc.b_mb_tree = x4->mbtree;
  209. if (x4->direct_pred >= 0)
  210. x4->params.analyse.i_direct_mv_pred = x4->direct_pred;
  211. if (x4->fast_pskip >= 0)
  212. x4->params.analyse.b_fast_pskip = x4->fast_pskip;
  213. if (x4->mixed_refs >= 0)
  214. x4->params.analyse.b_mixed_references = x4->mixed_refs;
  215. if (x4->b_bias != INT_MIN)
  216. x4->params.i_bframe_bias = x4->b_bias;
  217. if (x4->cplxblur >= 0)
  218. x4->params.rc.f_complexity_blur = x4->cplxblur;
  219. x4->params.i_bframe = avctx->max_b_frames;
  220. /* cabac is not included in AVS JiZhun Profile */
  221. x4->params.b_cabac = 0;
  222. x4->params.i_bframe_adaptive = avctx->b_frame_strategy;
  223. avctx->has_b_frames = !!avctx->max_b_frames;
  224. /* AVS doesn't allow B picture as reference */
  225. /* The max allowed reference frame number of B is 2 */
  226. x4->params.i_keyint_min = avctx->keyint_min;
  227. if (x4->params.i_keyint_min > x4->params.i_keyint_max)
  228. x4->params.i_keyint_min = x4->params.i_keyint_max;
  229. x4->params.i_scenecut_threshold = avctx->scenechange_threshold;
  230. // x4->params.b_deblocking_filter = avctx->flags & CODEC_FLAG_LOOP_FILTER;
  231. x4->params.rc.i_qp_min = avctx->qmin;
  232. x4->params.rc.i_qp_max = avctx->qmax;
  233. x4->params.rc.i_qp_step = avctx->max_qdiff;
  234. x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
  235. x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
  236. x4->params.i_frame_reference = avctx->refs;
  237. x4->params.i_width = avctx->width;
  238. x4->params.i_height = avctx->height;
  239. x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
  240. x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
  241. /* This is only used for counting the fps */
  242. x4->params.i_fps_num = avctx->time_base.den;
  243. x4->params.i_fps_den = avctx->time_base.num;
  244. x4->params.analyse.inter = XAVS_ANALYSE_I8x8 |XAVS_ANALYSE_PSUB16x16| XAVS_ANALYSE_BSUB16x16;
  245. switch (avctx->me_method) {
  246. case ME_EPZS:
  247. x4->params.analyse.i_me_method = XAVS_ME_DIA;
  248. break;
  249. case ME_HEX:
  250. x4->params.analyse.i_me_method = XAVS_ME_HEX;
  251. break;
  252. case ME_UMH:
  253. x4->params.analyse.i_me_method = XAVS_ME_UMH;
  254. break;
  255. case ME_FULL:
  256. x4->params.analyse.i_me_method = XAVS_ME_ESA;
  257. break;
  258. case ME_TESA:
  259. x4->params.analyse.i_me_method = XAVS_ME_TESA;
  260. break;
  261. default:
  262. x4->params.analyse.i_me_method = XAVS_ME_HEX;
  263. }
  264. x4->params.analyse.i_me_range = avctx->me_range;
  265. x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
  266. x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
  267. /* AVS P2 only enables 8x8 transform */
  268. x4->params.analyse.b_transform_8x8 = 1; //avctx->flags2 & CODEC_FLAG2_8X8DCT;
  269. x4->params.analyse.i_trellis = avctx->trellis;
  270. x4->params.analyse.i_noise_reduction = avctx->noise_reduction;
  271. if (avctx->level > 0)
  272. x4->params.i_level_idc = avctx->level;
  273. x4->params.rc.f_rate_tolerance =
  274. (float)avctx->bit_rate_tolerance/avctx->bit_rate;
  275. if ((avctx->rc_buffer_size) &&
  276. (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
  277. x4->params.rc.f_vbv_buffer_init =
  278. (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
  279. } else
  280. x4->params.rc.f_vbv_buffer_init = 0.9;
  281. /* TAG:do we have MB tree RC method */
  282. /* what is the RC method we are now using? Default NO */
  283. x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
  284. x4->params.rc.f_pb_factor = avctx->b_quant_factor;
  285. x4->params.analyse.i_chroma_qp_offset = avctx->chromaoffset;
  286. x4->params.analyse.b_psnr = avctx->flags & CODEC_FLAG_PSNR;
  287. x4->params.i_log_level = XAVS_LOG_DEBUG;
  288. x4->params.i_threads = avctx->thread_count;
  289. x4->params.b_interlaced = avctx->flags & CODEC_FLAG_INTERLACED_DCT;
  290. if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER)
  291. x4->params.b_repeat_headers = 0;
  292. x4->enc = xavs_encoder_open(&x4->params);
  293. if (!x4->enc)
  294. return -1;
  295. if (!(x4->pts_buffer = av_mallocz((avctx->max_b_frames+1) * sizeof(*x4->pts_buffer))))
  296. return AVERROR(ENOMEM);
  297. avctx->coded_frame = &x4->out_pic;
  298. /* TAG: Do we have GLOBAL HEADER in AVS */
  299. /* We Have PPS and SPS in AVS */
  300. if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {
  301. xavs_nal_t *nal;
  302. int nnal, s, i, size;
  303. uint8_t *p;
  304. s = xavs_encoder_headers(x4->enc, &nal, &nnal);
  305. avctx->extradata = p = av_malloc(s);
  306. for (i = 0; i < nnal; i++) {
  307. /* Don't put the SEI in extradata. */
  308. if (nal[i].i_type == NAL_SEI) {
  309. x4->sei = av_malloc( 5 + nal[i].i_payload * 4 / 3 );
  310. if (xavs_nal_encode(x4->sei, &x4->sei_size, 1, nal + i) < 0)
  311. return -1;
  312. continue;
  313. }
  314. size = xavs_nal_encode(p, &s, 1, nal + i);
  315. if (size < 0)
  316. return -1;
  317. p += size;
  318. }
  319. avctx->extradata_size = p - avctx->extradata;
  320. }
  321. return 0;
  322. }
  323. #define OFFSET(x) offsetof(XavsContext, x)
  324. #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
  325. static const AVOption options[] = {
  326. { "crf", "Select the quality for constant quality mode", OFFSET(crf), AV_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE },
  327. { "qp", "Constant quantization parameter rate control method",OFFSET(cqp), AV_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE },
  328. { "b-bias", "Influences how often B-frames are used", OFFSET(b_bias), AV_OPT_TYPE_INT, {INT_MIN}, INT_MIN, INT_MAX, VE },
  329. { "cplxblur", "Reduce fluctuations in QP (before curve compression)", OFFSET(cplxblur), AV_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE},
  330. { "direct-pred", "Direct MV prediction mode", OFFSET(direct_pred), AV_OPT_TYPE_INT, {-1 }, -1, INT_MAX, VE, "direct-pred" },
  331. { "none", NULL, 0, AV_OPT_TYPE_CONST, { XAVS_DIRECT_PRED_NONE }, 0, 0, VE, "direct-pred" },
  332. { "spatial", NULL, 0, AV_OPT_TYPE_CONST, { XAVS_DIRECT_PRED_SPATIAL }, 0, 0, VE, "direct-pred" },
  333. { "temporal", NULL, 0, AV_OPT_TYPE_CONST, { XAVS_DIRECT_PRED_TEMPORAL }, 0, 0, VE, "direct-pred" },
  334. { "auto", NULL, 0, AV_OPT_TYPE_CONST, { XAVS_DIRECT_PRED_AUTO }, 0, 0, VE, "direct-pred" },
  335. { "aud", "Use access unit delimiters.", OFFSET(aud), AV_OPT_TYPE_INT, {-1 }, -1, 1, VE},
  336. { "mbtree", "Use macroblock tree ratecontrol.", OFFSET(mbtree), AV_OPT_TYPE_INT, {-1 }, -1, 1, VE},
  337. { "mixed-refs", "One reference per partition, as opposed to one reference per macroblock", OFFSET(mixed_refs), AV_OPT_TYPE_INT, {-1}, -1, 1, VE },
  338. { "fast-pskip", NULL, OFFSET(fast_pskip), AV_OPT_TYPE_INT, {-1 }, -1, 1, VE},
  339. { NULL },
  340. };
  341. static const AVClass class = {
  342. .class_name = "libxavs",
  343. .item_name = av_default_item_name,
  344. .option = options,
  345. .version = LIBAVUTIL_VERSION_INT,
  346. };
  347. static const AVCodecDefault xavs_defaults[] = {
  348. { "b", "0" },
  349. { NULL },
  350. };
  351. AVCodec ff_libxavs_encoder = {
  352. .name = "libxavs",
  353. .type = AVMEDIA_TYPE_VIDEO,
  354. .id = CODEC_ID_CAVS,
  355. .priv_data_size = sizeof(XavsContext),
  356. .init = XAVS_init,
  357. .encode2 = XAVS_frame,
  358. .close = XAVS_close,
  359. .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
  360. .pix_fmts = (const enum PixelFormat[]) { PIX_FMT_YUV420P, PIX_FMT_NONE },
  361. .long_name = NULL_IF_CONFIG_SMALL("libxavs - the Chinese Audio Video Standard Encoder"),
  362. .priv_class = &class,
  363. .defaults = xavs_defaults,
  364. };